Jev vs GPT-5.6 Terra: cost per decision
Terra is the step up teams reach for when Luna starts getting classifications wrong. It bills six times Luna's output rate, which is the half of the bill Jev does not have at all.
| Scenario | Jev 1.13 | GPT-5.6 Terra | Ratio |
|---|---|---|---|
Light 10k items/day, 300-token state, 2 questions | $5.29 | $360.00 | 68x |
Production 100k items/day, 500-token state, 4 questions | $93.24 | $6,240 | 67x |
High volume 2M items/day, 300-token state, 1 question | $856.80 | $48,000 | 56x |
Monthly, 30 days. Both sides answer all questions for an item in a single call. GPT-5.6 Terra additionally pays for its output tokens at $12/M; Jev 1.13 has no output charge. List prices at standard tier, excluding caching and batch discounts.
Where GPT-5.6 Terra still wins
Terra reasons over ambiguity better than a typed question can express. If your labels overlap and the boundary is genuinely a judgement call, that judgement is what you are paying for.
The honest caveat
This compares price, not accuracy. A model that is 100x cheaper and wrong 5% more often can easily be the more expensive choice once you price the mistakes. Jev returns a calibrated probability with every answer, so the practical move is to route the confident cases automatically and send the rest to GPT-5.6 Terra or a human — which makes this a routing decision rather than a replacement decision.